 |
 |
Ben and others, Please see inline responses below. On Sun, Jul 22, 2012 at 1:03 AM, Ben Becker <[log in to unmask]> wrote: > Dear SPMers, > > we recently implemented the analysis of functional connectivity in our lab using standard (standard SPM 8 PPI & gPPI from Donald McLaren (http://www.ncbi.nlm.nih.gov/pubmed?term=PMID%3A%2022484411). In this context we particularly discussed the way to validely define seed regions. A recent scan paper from O’Reilly et al. (http://www.ncbi.nlm.nih.gov/pubmed?term=PMID%3A%2022569188) addresses this issue & concludes that defining seed regions based on (a) strongest task effect in a group analysis & (b) selecting each participants voxel with the strongest effect in a predefined anatomical volume of interest does not represent a case circularity “as long as we model the main effect of task when we run the PPI analysis. In this case the PPI will only detect functional connectivity effects over & above (orthogonal to) the main effect of task (p. 606, legend to figure 1)” >>> Thanks for pointing out an excellent paper describing the PPI method and interpretation. It provides an nice overview of the more technical paper that describes gPPI. > > My questions: > > - sPPI: Does the standard SPM8 PPI analysis account for this? However, it only allows to account for two task conditions – what about more complex tasks employing more task conditions? Does the main effect have to be specified in a separate model? (& perhaps serve as a mask during the analysis of PPI Effects?) >>> Since the standard SPM8 PPI analysis models your two conditions as A-B, then you are only modelling the main effects IF there are only two conditions. If you have condition A, condition B, and fixation, then you have 3 conditions. Thus, modeling the main effects in SPM8 isn't done except for very limited designs. The main task effects should be included as regressors as done with gPPI. > > - gPPI: gPPI allows to adjust the data for the Omnibus F-test for PPI Analyses – does this correspond to modeling the main effect of task? >>>>> No. This is done for the purpose of deconvolution to remove the effects of motion and any other noise covariates that were specified in the GLM that aren't due to your task regressors. In my paper, these regressors are G. In SPM, they are labelled in SPM.Sess.c. Hope this helps. Let me know if you have any other questions. > > Thanks in advance & best regards > > Ben